diplomat33
Average guy who loves autonomous vehicles
Nice to see the Waymo-Geely vehicle on display at CES. I am hoping we see deployment of the Geely vehicle this year.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Waymo
- Thread starter Daniel in SD
- Start date
Doggydogworld
Active Member
Spinning lidars always have these concentric lines on the ground.
Solid state and MEMS lidars vary, some can adjust elevation and azimuth independently to produce a more rectangular-looking pattern. In all cases the "lines" are just a series of points that are very close together. It's generally easier for lidar engineers to increase point density horizontally than vertically (especially if you reduce frame rate for a nice PR image).
The original Velodyne HDL-64e "spinning KFC bucket" resolution was 64 x 1024 (V x H) at 20 fps, but 64 x 4096 if you slowed it down to 5 fps. Today's high end lidars have even more resolution. When you display a couple thousand adjacent points on a computer monitor they mush together into a line.
Solid state and MEMS lidars vary, some can adjust elevation and azimuth independently to produce a more rectangular-looking pattern. In all cases the "lines" are just a series of points that are very close together. It's generally easier for lidar engineers to increase point density horizontally than vertically (especially if you reduce frame rate for a nice PR image).
The original Velodyne HDL-64e "spinning KFC bucket" resolution was 64 x 1024 (V x H) at 20 fps, but 64 x 4096 if you slowed it down to 5 fps. Today's high end lidars have even more resolution. When you display a couple thousand adjacent points on a computer monitor they mush together into a line.
powertoold
Active Member
Solid concentric circles are an augmented visual aid on top of the actual LIDAR points.
Most visual LIDAR representations are augmented for human consumption. Otherwise, you'd just get a bunch of blinking points.
This is just my common sense intuition, so feel free to find some raw LIDAR point clouds to compare.
diplomat33
Average guy who loves autonomous vehicles
Doggydogworld
Active Member
Lidar point clouds are just massive strings of numbers.
If you visually represent 2000 horizontal points in a 1280x800 image you get a line. It's unavoidable when you have more points than pixels.
Raw lidar data are just numbers that represent X,Y,Z and other data. The same raw data can be represented graphically in the form of points in 3d space. The concentric circular pattern is just a result of how that particular lidar hardware scans.
A rotating lidar would create a circular pattern, here is an example using Velodyne HDL 32E surround lidar sensor sample data. Link provided below so you can play with it yourself. To the right is the raw lidar data which are just numbers, and to the left is the graphical visualization of the raw data in 3d space. The concentric circle is not an augmented visual aid.
The lidar on an iphone shoots out a grid of points
Raw lidar data from Velodyne HDL32 sensor here and Veloview software to view it.
Last edited:
diplomat33
Average guy who loves autonomous vehicles
Video of the Waymo-Geely vehicle that shows the trunk space, sensors and interior.
More pics here:
thelastdriverlicenseholder.com
According to this article, the vehicle is in production and will start public road testing "soon".
More pics here:
Waymo’s New Robotaxi Without Steering Wheel
At CES 2023 in Las Vegas, Waymo presented the prototype of the new platform in addition to the previously known autonomous vehicles, some of which are already historic. It is based on the battery-e…
thelastdriverlicenseholder.com
According to this article, the vehicle is in production and will start public road testing "soon".
the concentric circles phenomenon seems so obvious to me. @powertoold, you are wrong, sorry.
You have a point source of light (the lidar unit), and it's aimed toward the front, slightly downward. So the circular lines represent equidistance from your point source. If your road is normal (perpendicular) to the light beams, then you would see an evenly spaced matrix of dots (or in cases of high resolution, a solid shape) visualized. But that's an impossibly steep road. Most roads are going to be flat, which means more parallel to the beams of light. So the less steep the road, the more space between vertical light beams. If the road goes downhill, the spacing between the horizontal scan lines increases. If the road goes fully parallel, then the spacing between scan lines becomes infinite.
To minimize the spacing between concentric circles, you'd have to increase the vertical resolution of the lidar. But it really isn't necessary, as the spacing works similarly to topographical maps. You can see what is flat and what is steep based on the density of the horizontal/curved lines.
The "lines" in the visualization are just densely packed points, and the resolution of the lidar is higher than the pixel density of the display, so it appears to be a solid line. In a lower-res lidar where the display pixel density is not exceeded, you'd see dotted lines instead, and even more spacing between concentric lines.
You have a point source of light (the lidar unit), and it's aimed toward the front, slightly downward. So the circular lines represent equidistance from your point source. If your road is normal (perpendicular) to the light beams, then you would see an evenly spaced matrix of dots (or in cases of high resolution, a solid shape) visualized. But that's an impossibly steep road. Most roads are going to be flat, which means more parallel to the beams of light. So the less steep the road, the more space between vertical light beams. If the road goes downhill, the spacing between the horizontal scan lines increases. If the road goes fully parallel, then the spacing between scan lines becomes infinite.
To minimize the spacing between concentric circles, you'd have to increase the vertical resolution of the lidar. But it really isn't necessary, as the spacing works similarly to topographical maps. You can see what is flat and what is steep based on the density of the horizontal/curved lines.
The "lines" in the visualization are just densely packed points, and the resolution of the lidar is higher than the pixel density of the display, so it appears to be a solid line. In a lower-res lidar where the display pixel density is not exceeded, you'd see dotted lines instead, and even more spacing between concentric lines.
powertoold
Active Member
Thanks for your analysis, but how would you explain the transition from the dense single points to the solid concentric lines in this picture that diplomat posted?
Separately, how would you explain the single points behind the solid lines:
diplomat33
Average guy who loves autonomous vehicles
As @nvx1977 mentioned, there are likely different point clouds from different lidar sources in that image, one from the long range spinning lidar on the roof, one from the short range lidar on the bumpers. The solid concentric lines are from the long range spinning lidar on the roof. The dense single points we see closer to the vehicle are from the short range lidar on the bumpers.
Again, different point clouds from different lidars.
powertoold
Active Member
I know, but at first you were saying that it's so obvious I was wrong, but now it seems you aren't sure.
Why would there be multiple data feeds? We are just talking about the raw point cloud data.
I don't know. To me, it's very obvious you guys aren't thinking deeply enough about what you're seeing in these pictures. I'm not saying I'm right since I'm just using my common sense, but you aren't explaining all the contradictory details in the pictures.
powertoold
Active Member
Dude, I literally showed you a close up of hundreds of single points behind the solid concentric circles. This is the second time I posted this detail. Are you saying the lidar on the roof produces points AND lines?
diplomat33
Average guy who loves autonomous vehicles
You are not using common sense.
There are multiple data feeds because there are multiple lidar sources that are different resolutions and different wavelengths.
Yes, you can think of it as one point cloud but it is composed of point clouds from different sources.
I want to be nice but how you are still not getting it? It has explained to you many times now. The lidar on the roof does not produces points and lines. It only produces points BUT the points are dense enough and close enough that they look like lines on the image.
The "lines" are points very close together from the spinning lidar on the roof. The single points behind the solid concentric lines are also points but from a different lidar. The reason there are very few points behind the lines is because they are from the short range lidar so there are very few points that far away.
because the concentric pattern is such a defining quality of a spinning lidar, and it's also easy to explain. Not sure why you remain so dismissive of this and continue to think your "visual augmentation" hypothesis holds more water, esp after numerous other lidar visualizations have been posted showing the same phenomenon.
powertoold
Active Member
Are you guys intentionally ignoring my posts / pictures? Or are you blind?
Behind the solid concentric circles are hundreds of single points that are grouped in a concentric pattern. Are you saying there's two separate spinning lidars in the roof that generate these two separate data streams?
Also, not sure when you joined this discussion, but we aren't talking about lidar visualizations in general. We are talking about raw lidar point clouds.
diplomat33
Average guy who loves autonomous vehicles
Solid concentric lines = spinning lidar on the roof
Hundreds of single points behind the solid concentric lines = perimeter lidar on the bumpers.
I don't know how much simpler I can explain it.
Last edited:
powertoold
Active Member
Yes, you are clear, but you're purposely avoiding explaining why the solid lines are visually expressed differently than all the actual lidar points.
A lidar point cloud is made up of points, not lines. Not only are the lines in the picture solid, but they're also transparent, showing the points underneath.
My explanation is that the solid lines are a visual aid derived from the points. The solid lines makes it easier for a human to see what's in the picture.
Similar threads
- Article
